Thursday, August 21, 2008

Big Moves on Low Volume

With volume having drifted back down since the passing of July's stormy lows, we now find ourselves in the midst of the summer doldrums. On the other hand, August has certainly seen its share of large price swings.

As promised earlier in the week, this study takes a quick look at what typically happens the day after a "big move on low volume."

The conventional wisdom is to fade such moves -- if the market is significantly higher, the trader should go short, and vice-versa. Is this correct? As we will see, data is sparse, but anecdotally -- yes and no. Not surprisingly, the answer depends on how one defines big and low.

The analysis spans ten years of S&P 500 data (2,520 trading days), using the 'SPY' Exchange Traded Fund as a frictionless proxy. The table below presents next day results assuming the trader goes against the conventional wisdom in both directions, seeking a trend continuation trade. First, however, here are some definitions:

  • "Big Move" - A daily price change that is 1.8 to 2.2-plus standard deviations above the preceding one-year average.

  • "Low Volume" - Closing volume that is up to 20% or more below the preceding one-year average.
  • These events are rare, limiting the study's statistical significance. In fact, no such trades have actually triggered this month (although we have come close a few times, stimulating the timing of this piece). Also, not seen here is how market behaviour changes over time. Nonetheless, we can still anecdotally test our query.

  • Contrary to conventional wisdom, for moderately low volume between 80% and 100% of average, there is a strong tendency for continuation.

  • Although the anecdotal odds of continuation trade success are quite high in both directions, a few large losses on the short side completely mitigated wins there except for the largest of pullbacks.

  • As volume gets increasingly thin (below 80% of average), consistent with conventional wisdom, the anecdotal odds again favor a fadable reversal.

For more analysis of volume indicated trading, see this excellent series by Quantifiable Edges. As interesting as the results of this series are, it is also instructive in showing how different conclusions may be drawn by slightly altering parameters or approach. This highlights the importance of confirming the robustness of results in all back testing, certainly far beyond the scope of this piece.

Lastly, estimating end-of-day volume to see if it will qualify as "low" while the market is still open can be problematic, since: a) by definition closing volume is not yet posted; b) volume is typically noisy; and, c) the distribution of volume over the course of the day is nonlinear. You may be able to get around these problems by using the divisors provided in this Market Rewind post.

In the future, I'll also look at post high-volume days. As we will see, there is much more information and compelling profit potential following those days. Can you guess what the title of that article will be?

Never Investment Advice

No comments: